How to Build a Recruiting Agent in CogniAgent (Step by Step)

Jun 15, 2026
5 min read

Hiring is repetitive. Every candidate gets the same greeting, the same questions, the same follow-up. That’s exactly the kind of work an AI agent should handle – so your team can focus on the people who actually make it to the next round.

This guide walks you through building a full recruiting agent in CogniAgent’s Conversation Flows: from greeting the candidate to logging their answers in Notion and sending a summary to the hiring manager via Gmail.

What You’re Building

A conversational recruiting agent that:

  • Greets candidates and confirms the role they’re applying for
  • Conducts a structured screening interview
  • Logs all answers to Notion automatically
  • Sends a candidate summary to the hiring manager via Gmail

No forms. No scheduling. No manual data entry.

Step 1: Create a New Conversation Flow

Go to Conversations -> New Flow 

Open Flow Settings (gear icon) and configure:

General tab:

  • Flow Name: Recruiting Agent
  • Execution Mode: Responder – the candidate writes first

Global Context:

You are a professional recruiting assistant for Acme LTD, a fast-growing SaaS company. You are conducting a screening interview for the Growth Marketing Manager position.

Tone: warm, professional, concise.

Rules:

– Ask one question at a time

– Never rush the candidate

– Do not evaluate or judge answers during the conversation

– Do not ask for information already provided

– If the candidate goes off-topic, gently redirect

– Never reveal internal hiring criteria

– Always refer to the company as “Acme LTD”

Global Context is injected into every actor’s system prompt automatically – you only write it once.

Step 2: Set Up the Greetings Actor

Click Add Actor and name it Greetings Actor.

Activation Condition:

Candidate sends the first message through the channel

Basic tab:

Context:

You work for Acme LTD, a fast-growing SaaS company.

Instructions:

Greet the candidate warmly, give a one-sentence intro about Acme LTD, and confirm they are here to apply for the Growth Marketing Manager position.

Definition of Done:

Candidate has confirmed they are applying for the Growth Marketing Manager position.

Inherit Context: ✅ Global flow context ✅ Collected information so far

The Definition of Done is critical — without it, the actor won’t know when to hand off to the next stage and will keep looping.

Set:
The candidate has confirmed they are applying for the Growth Marketing Manager position.

Focus Mode:

Controls how the actor handles topic changes mid-conversation. Choose “Persistent”.

Step 3: Set Up the Screening Actor

Add a second actor: Screening Actor.

Activation Condition:

Greeting is complete and candidate is ready for screening

Basic tab:

Focus Mode: Strict — forces the actor to follow your steps in order instead of improvising.

Inherit Context: ✅ Global flow context ✅ Collected information so far

Advanced tab — Add these steps in order:

Step 1 — Ask Question

Tell me a bit about yourself and your background in growth marketing.

Variable: intro | Max attempts: 3 | ✅ Skip if already collected

Step 2 — Ask Question

What growth channels have you owned end-to-end — paid, SEO, email, partnerships?

Variable: channels_owned | Max attempts: 3 | ✅ Skip if already collected

Step 3 — Ask Question

Walk me through a campaign you ran that had a measurable impact on revenue or user growth.

Variable: best_campaign | Max attempts: 3 | ✅ Skip if already collected

Step 4 — Ask Question

How do you decide where to focus when you have limited budget and multiple channels to test?

Variable: prioritization | Max attempts: 3 | ✅ Skip if already collected

Step 5 — Ask Question

What tools do you work with daily — analytics, automation, ads?

Variable: tools_stack | Max attempts: 3 | ✅ Skip if already collected

Step 6 — Ask Question

What is your expected monthly salary and when could you start?

Variable: salary_and_availability | Max attempts: 3 | ✅ Optional question

Step 7 — Static Message

Thank you — your answers give us a great picture of your experience. The Acme LTD team will review your application and get back to you within 3–5 business days. Good luck! 🙌

Step 8 — Act

After all questions are collected, add an Act step. This is where the actor executes actions based on the interview data — no separate workflow node needed.

Step 9 — Static Message

Thank you — your answers give us a great picture of your experience. 

The Acme LTD team will review your application and get back to you 

within 3–5 business days. Good luck! 🙌

Step 10 — End

Note: Notion and Gmail must be connected under Capabilities -> Integrations inside the Screening Actor before the Act step can execute. Without active integrations the actor has no tools to work with.

Step 4: Connect Notion and Gmail as Capabilities

Inside the Screening Actor, go to Capabilities -> Integrations and follow the instructions to connect Notion and Gmail.

Step 5: Configure the Channel

Go to Flow Settings -> Channels and connect your channel – Web Widget, Gmail, SMS, WhatsApp, Telegram, etc. This is where candidates will reach the agent and start the screening process.

Step 7: Test Before Deploying

Click Test at the bottom of the canvas. Use the debug panel to verify:

  • ✅ ROUTING → Greetings Actor fires on first message
  • ✅ ACTOR DONE() triggers after candidate confirms the role
  • ✅ ROUTING → Screening Actor fires after greeting completes
  • ✅ Steps run in order (Strict mode)
  • ✅ Notion page is created after interview ends
  • ✅ Gmail summary arrives in hiring manager’s inbox

Once all checks pass — hit Deploy.

The same pattern works for any role — just swap the screening questions and update the Global Context. Follow other guides on building with CogniAgent to explore more use cases — from customer support agents to lead qualification flows and automated onboarding.